New binding mode to TNF-alpha revealed by ubiquitin-based artificial binding protein

Abstract

A variety of approaches have been employed to generate binding proteins from non-antibody scaffolds. Utilizing a beta-sheet of the human ubiquitin for paratope creation we obtained binding proteins against tumor necrosis factor (TNF)-alpha. The bioactive form of this validated pharmacological target protein is a non-covalently linked homo-trimer. This structural feature leads to the observation of a certain heterogeneity concerning the binding mode of TNF-alpha binding molecules, for instance in terms of monomer/trimer specificity. We analyzed a ubiquitin-based TNF-alpha binder, selected by ribosome display, with a particular focus on its mode of interaction. Using enzyme-linked immunosorbent assays, specific binding to TNF-alpha with nanomolar affinity was observed. In isothermal titration calorimetry we obtained comparable results regarding the affinity and detected an exothermic reaction with one ubiquitin-derived binding molecule binding one TNF-alpha trimer. Using NMR spectroscopy and other analytical methods the 1:3 stoichiometry could be confirmed. Detailed binding analysis showed that the interaction is affected by the detergent Tween-20. Previously, this phenomenon was reported only for one other type of alternative scaffold-derived binding proteins--designed ankyrin repeat proteins--without further investigation. As demonstrated by size exclusion chromatography and NMR spectroscopy, the presence of the detergent increases the association rate significantly. Since the special architecture of TNF-alpha is known to be modulated by detergents, the access to the recognized epitope is indicated to be restricted by conformational transitions within the target protein. Our results suggest that the ubiquitin-derived binding protein targets a new epitope on TNF-alpha, which differs from the epitopes recognized by TNF-alpha neutralizing antibodies.

Figure 1. Amino acid positions chosen for randomization in library construction.

Based on the scaffold ubiquitin (shown here in cartoon representation), in particular with an F45W substitution, a library for the selection of artificial binding proteins was generated. For this purpose the six surface-exposed amino acid residues K6, L8, R42, I44, H68 and V70 (highlighted in red), located in the beta-sheet region of the scaffold, were chosen to be randomized for library construction. After in vitro selection against TNF-alpha, in the ubiquitin variant named 10F the residues D58 and Y59 (highlighted in blue) were found deleted. This figure was generated using pdb entry 1UBI and the software PyMOL version 0.99rc6 (DeLano Scientific LLC, South San Francisco, CA).

Figure 2. ELISA analysis of the TNF-alpha binding ubiquitin variant 10F.

(A) Specificity of ubiquitin variant 10F. The interaction of the purified, His₆-tagged binder with different immobilized target proteins (500 ng per well, human serum was used without dilution) is shown. (NGF: nerve growth factor, nPAC-1Rs: short splice form of N-terminal domain of human PACAP-receptor 1) (B) Characterization of the specificity by competition of the interaction of TNF-alpha and 10F. Biotinylated TNF-alpha (1 µM), preincubated for 1 h with competitor (etanercept: TNF-receptor 2 dimerized by fusion to IgG F_c), was incubated with immobilized 10F (100 ng per well). Detection by avidin-peroxidase demonstrates competition by etanercept. (C) Concentration-dependent competition ELISA. 10F (100 nM, His₆-tagged), preincubated with different concentrations of TNF-alpha, was incubated with immobilized TNF-alpha (150 ng per well). An IC₅₀ value of 289±53 nM was determined.

Figure 3. Isothermal titration calorimetry of TNF-alpha and the ubiquitin variant 10F.

10F at 135 µM was injected into the measurement cell containing TNF-alpha at 35.6 µM. The titration was carried out at 25°C. Experiments were conducted in duplicate. The upper panel shows the raw data and the lower panel shows the integrated enthalpy data corrected for dilution heat of a single experiment. Fitting the data to a single-site binding model (solid line), we determined K_D = 180±50 nM, stoichiometry factor N = 0.35±0.04, ΔH = −57.3±2.5 kJ mol⁻¹ and TΔS = −18.8±3.2 kJ mol⁻¹.

Figure 4. NMR analysis of the titration of 10F with TNF-alpha.

¹H-¹⁵N fHSQC spectra of ¹⁵N-labeled ubiquitin variant 10F in presence of increasing concentration of unlabeled TNF-alpha were recorded in PBS, 1 mM EDTA pH 7.4 with 0.05% Tween-20. The relative intensity decrease of two amid resonances was plotted against molar ratio of both proteins. As determined by the intersection point of the linear fits of the initial part and the plateau part of the curve 2.88±0.18 TNF-alpha monomers are bound per 10F molecule.

Figure 5. Size exclusion chromatography (SEC) of the complex of TNF-alpha and the ubiquitin variant 10F.

(A) SEC in presence of detergent Tween-20. Samples of fluorescein-labeled 10F (10 µM), TNF-alpha (60 µM), and a mixture of both, were pre-incubated for 10 min at 4°C and loaded on a Superose 12 size-exclusion column. Experiments were performed with sample and running buffer, both containing 0.05% (v/v) Tween-20, and detection wavelengths of 280 nm (blue line) and 495 nm (red dashed line). Note that there is no significant shift of the TNF-alpha elution volume when comparing the elution profiles of TNF-alpha alone and the mixture. Fluorescein itself did not co-elute with TNF-alpha (data not shown). (B) SEC in absence of detergent Tween-20. For analysis of Tween-20 independent complex formation, a detergent-free mixture of unlabeled 10F (10 µM) and TNF-alpha (60 µM) was incubated at 4°C and aliquots were analyzed as described before using running buffer without Tween-20. Note that the area of the 10F peak (16.5 ml) decreases during prolonged incubation. Dashed lines mark the borders of the fraction used for SDS-PAGE analysis. (C) SDS-PAGE analysis of the detergent-free SEC complex fractions. Aliquots of appropriate SEC fractions were analyzed by SDS-PAGE (4–12% gradient gel) followed by Coomassie staining. M: Fermentas PageRuler Unstained, 1: control TNF-alpha (2 µg), 2: control 10F (2 µg), 3: 0.17 h preincubation, 4: 1 h, 5: 2 h*, 6: 6 h, 7: 12 h, 8: 24 h*, 9: 44 h. *To reduce complexity chromatograms were not included in (B).

Figure 6. Circular dichroism analysis of ubiquitin variants.

The far-UV CD spectra of the native proteins 10F (open circles), ubiquitin F45W (open triangles), and 10F_DY (open squares) were recorded in 10 mM KH₂PO₄ pH 7.4 at 20°C. For analysis of denatured proteins (closed symbols, 10F_DY not shown), samples were measured in 6 M guanidine hydrochloride. All spectra are buffer corrected.

Figure 7. 2D-NMR characterization of the scaffold ubiquitin F45W and the variant 10F.

¹H-¹⁵N fHSQC spectra of the scaffold ubiquitin F45W (black, 900 µM) and its variant 10F (red, 400 µM) were acquired on an 800 MHz NMR spectrometer at 25°C. Both proteins were dissolved in PBS, 1 mM EDTA at pH 7.4. Superimposition of the spectra shows significant deviation of chemical shifts for almost all amino acids including the conserved ones. Note, that most residues between 130 ppm and 137 ppm are aliased along the ¹⁵N-dimension.